import spacy
import gradio as gr
from spacy.pipeline import EntityRuler
from spacy import displacy
import jsonlines
from spacy.cli import download
download('en_core_web_sm')
nlp = spacy.load('en_core_web_sm')

# Create list with entity labels from jsonl file
with jsonlines.open("skill_patterns.jsonl") as f:
    created_entities = [line['label'].upper() for line in f.iter()]
    
def extract_text_from_word(txt):
    '''Opens en reads in a .doc or .docx file from path'''  
    return txt.replace('\n', ' ').replace('\t', ' ').lower()

def add_newruler_to_pipeline(skill_pattern_path):
    '''Reads in all created patterns from a JSONL file and adds it to the pipeline after PARSER and before NER'''
    
#     new_ruler = EntityRuler(nlp).from_disk(skill_pattern_path)
    ruler=nlp.add_pipe("entity_ruler",after='parser')
    ruler.from_disk(skill_pattern_path)  # loads patterns only
    
def create_skill_set(doc):
    '''Create a set of the extracted skill entities of a doc'''
    
    return set([ent.label_.upper()[6:] for ent in doc.ents if 'skill' in ent.label_.lower()])

def create_skillset_dict(resume_names, resume_texts):
    '''Create a dictionary containing a set of the extracted skills. Name is key, matching skillset is value'''
    skillsets = [create_skill_set(resume_text) for resume_text in resume_texts]
    return dict(zip(resume_names, skillsets))
    
def match_skills(vacature_set, cv_set, resume_name):
    '''Get intersection of resume skills and job offer skills and return match percentage'''
    
    if len(vacature_set) < 1:
        print('could not extract skills from job offer text')   
    else:
        pct_match = round(len(vacature_set.intersection(cv_set[resume_name])) / len(vacature_set) * 100, 0)
        print(resume_name + " has a {}% skill match on this job offer".format(pct_match))
        print('Required skills: {} '.format(vacature_set))
        print('Matched skills: {} \n'.format(vacature_set.intersection(cv_set[resume_name])))
        
        return (resume_name, pct_match)
    
add_newruler_to_pipeline("skill_patterns.jsonl")   
 
def match(CV,JD):
    resume_texts=[]
    resume_texts.append(nlp(CV))
    resume_names=['ABHI']
    skillset_dict = create_skillset_dict(resume_names, resume_texts)
    jd_skillset = create_skill_set(nlp(JD))
    match_pairs = [match_skills(jd_skillset, skillset_dict, name) for name in skillset_dict.keys()]
    if match_pairs[0]:
        return match_pairs[0][1]
    else:
        return "No matching skill set."    

exp=["Who is steve jobs?","What is coldplay?","What is a turing test?","What is the most interesting thing about our universe?","What are the most beautiful places on earth?"]

desc="A Machine Learning Based Resume Matcher, to compare Resumes with Job Descriptions. "

inp1=gr.inputs.Textbox(lines=10, placeholder=None, default="", label="Resume Details")
inp2=gr.inputs.Textbox(lines=10, placeholder=None, default="", label="Job Description")

out=gr.outputs.Textbox(type="auto",label="Match Score")

iface = gr.Interface(fn=match, inputs=[inp1,inp2], outputs=out,title="A Machine Learning Based Resume Matcher, to compare Resumes with Job Descriptions",article=desc,theme="huggingface",layout='vertical')
iface.launch(debug=True)